Skip to main content

Full text of "Chemical abundance ratios of galactic globular clusters from modelling integrated light spectroscopy"

See other formats


Mon. Not. R. Astron. Soc. 0Q0,[Tlfl3l(2010) Printed 4 November 2010 (MN WTtX style file v2.2) 



Chemical abundance ratios of galactic globular clusters from 
modelling integrated light spectroscopy 

Daniel Thomas, Jonas Johansson, Claudia Maraston 

Institute of Cosmology and Gravitation, University of Portsmouth, Dennis Sciama Building, Burnaby Road, Portsmouth, POl 3FX, UK 
SEPNET, South East Physics Network 

Accepted ... Received 20 October 2010 ; in original form 14 July 2010 

ABSTRACT 

In a companion paper we present new, flux-calibrated stellar population models of Lick 
absorption-line indices with variable element abundance ratios. The model includes a large 
variety of individual element variations, which allows the derivation of the abundances for 
the elements C, N, O, Mg, Ca, Ti, and Fe besides total metallicity and age. We use this 
model to obtain estimates of these quantities from integrated light spectroscopy of galactic 
globular clusters. We show that the model fits to a number of indices improve considerably 
when various variable element ratios are considered. The ages we derive agree well with the 
literature and are all consistent with the age of the universe within the measurement errors. 
There is a considerable scatter in the ages, though, and we overestimate the ages preferentially 
for the metal-rich globular c lusters. Our derived total metallicities agree generally very well 
with literature values on the Zinn & West! (119841) scale once corrected for a-enhancement, in 
particular for those cluster where the ages agree with the CMD ages. We tend to slightly un- 
derestimate the metallicity for those clusters where we overestimate the age, in line with the 
age-metallicity degeneracy. It turns out that the derivation of individual element abundance 
ratios is not reliable at an iron abundance [Fe/H] < — 1 dex where line strengths become 
weaker, while the [a/Fe] ratio is robust at all metallicities. The discussion of individual el- 
ement ratios focuses therefore on globular clusters with [Fe/H] > — 1 dex. We find general 
enhancement of light and a elements, as expected, with significant variations for some ele- 
ments. The elements O and Mg follow the same general enhancement with almost identical 
distributions of [O/Fe] and [Mg/Fe] . We obtain slightly lower [C/Fe] and very high [N/Fe] ra- 
tios, instead. This chemical anomaly, commonly attributed to self-enrichment, is well known 
in globular clusters from individual stellar spectroscopy. It is the first time that this pattern is 
obtained also from the integrated light. The a elements follow a pattern such that the heavier 
elements Ca and Ti are less enhanced. More specifically, the [Ca/Fe] and [Ti/Fe] ratios are 
lower than [O/Fe] and [Mg/Fe] by about 0.2 dex. Most interestingly this trend of element 
abundance with atomic number is also seen in recent determinations of element abundances 
in globular cluster and field stars of the Milky Way. This suggests that Type la supernovae 
contribute significantly to the enrichment of the heavier a elements as predicted by nucle- 
osynthesis calculations and galactic chemical evolution models. 

Key words: stars: abundances Galaxy: abundances globular clusters: general galaxies: 
formation galaxies: stellar content 



1 INTRODUCTION 

The abundances of a large variety of chemical elements can be de- 
rived from high-resolution spectroscopy of individual stars in the 
field and globular clusters of the Milky Way as well as nearby dwarf 
galax i es in the Local Group (e.g. , Mc Willi am! 1 9971; IC arretta et al. 
120051 : IPritzlet alj[2003 : iTolstov et alj|2009l : [Bensbv et alj|201(t) . 
This level of detail cannot be achieved for most galaxies and extra- 
galactic globular clusters, because the individual stars are not re- 
solved. Observations have to resort to integrated light spectroscopy, 



which is applicable to unresolved stellar populations. It allows us to 
study element abundances in distant galaxies and globular clusters, 
but is naturally more limited. Nearby globular clusters are the in- 
terface between these two extremes. They allow detailed chemical 
analyses from resolved stellar spectroscopy as well as the study of 
their integrated light. They are therefore vital for the calibration of 
stellar population models and integrated light analyses. 

The Lic k group have defined a set of 25 optical absorption- 
line indices dBurstein etal]|l984r . [Faber et alj|l985l : iGorgas et al.1 



Thomas, Johansson, Maraston 



1993| :IWorthev et alJl994l ; lTrager et al.ll998l ; IWorthev & Ottavianil 



19971) . the so-called Lick index system, that are by far the most 
commonly used in absorption-line analyses of old stellar popula- 
tions. These can be used, in principle, for the derivation of var- 
ious element abundance ratios. The bandpasses of Lick indices 
are relatively large with widths up to 50 A, which increases the 
signal-to-noise ratios but complicates their use for the derivation of 
individual element abundances. The first stellar population mod- 
els of Lick absorption-line indices with variable e lement abun- 
dance ratios have been published by Tho mas et alj j2003al 120041 
he reafter TMB/K models) base d on the index res ponse functions 
by iTripicco & Belli dl995h an d iKorn et al.1 (2005). In a compan- 
ion paper dThomas et alj|201ol . hereafter Paper I) we have updated 
these models (hereafter TMJ models) that are now flux calibrated 
thanks to the u s e of the newly computed index calibrations by 
Ijohansson etail d2010l) based on the fl ux-calibrated stellar library 
MILES dSanchez-Blazquez et alj2006I) . 

The lKorn et al.1 (2005J) model atmosphere calculations provide 
index response functions for the variation of the ten elements C, N, 
O, Mg, Na, Si, Ca, Ti, Fe, and Cr. Through additional features in 
the same part of the spectrum and modificatio ns of the index defi- 
nitions even mo re elements may be accessible dServen et al. 2005; 
iLee et ai]|2009h . Here we focus on those elements that can be best 
derived from the 25 Lick indices considered in the TMJ models. 
These are C, N, Mg, Ca, Ti, and Fe besides total metallicity Z/~R, 
age, and a/Fe ratio. 

We already used the iThomas et al.1 d2003al) code to derive 
abundances of nitrogen and calciu m for globular clusters and galax- 
ies. Through a simple approach in lThomas et ai] d2003al) we could 
show that galactic globular clusters must be significantly enhanced 
in nitrogen at fixed car bon abundance in ord er to reproduce the 
observed CN indices. In lThomas et al.1 d2003bh we derive calcium 
abund ances of galax i es. Su bsequent work has developed this fur- 
ther. IClemens et all d2006l) add C abun dances in their study of 
SDSS galaxies, and Kelson et al. (2006) deriv e N abundances of 
distant galaxies. iGraves & Schiavo n (2008) an d lSmithetal.ld2009h 
present the first full analyses of the abundances of C, N, Mg, Ca, 
and Fe in galaxies. In this paper we conduct the next step by 
adding the element titanium and derive the element abundance ra- 
tios [C/Fe], [N/Fe], [O/Fe], [Mg/Fe], [Ca/Fe], and [Ti/Fe] for galac- 
tic globular clusters. 

The paper is organised as follows. In Section [2] we describe 
the globular cluster data used. In Section [3] we introduce the new 
TMJ model and outline our method to derive element ratios. The 
main analysis is presented in Section [5] The results are discussed 
in Section[6] and the paper concludes with Section|7] 



2 GLOBULAR CLUSTER DATA 

Following our strategy for the TMB/K models, in Paper I we com- 
pare the model predictions with observational data of galactic glob- 
ular clusters, as the latter are the cl osest analogues of sim ple stel- 
lar populations in the real universe (Maraston et al. 2003). Key is 
that independent estimates of ages, metallicities, and element abun- 
dance ratios are available for the globular clusters of the Milky Way 
from deep photometry and high-resolution stellar spec t roscop y. 

The glob ular cluster samples ar e from lPuzia et al.l d2002l here- 
after P02) and lSchiavon et alJ d2005l hereafter S05). Critical for the 
integrated light spectroscop y is a representative sampling o f the un- 
derlying stellar population (Renzini 1998; Maraston 1998). To en- 
sure this P02 obtained several spectra with slightly offset pointings. 



In general three long-slit spectra were taken for each of the tar- 
get clusters, and the observing pattern was optimized to obtain one 
spectrum of the nuclear region and spectra of adjacent fields. Ex- 
posure times were adjusted according to the surface brightness of 
each globular cluster to reach a statistically secure luminosity sam- 
pling of the underlying stellar population. S05, instead, obtained 
each observation by drifting the spectrograph slit across the core 
diameter of the cluster. The telescope was positioned so as to off- 
set the slit from the cluster center by one core radius. A suitable 
trail rate was chosen to allow the slit to drift across the cluster core 
diameter during the typically 15 minute long exposure. 

We do not use the indices tabulated in P02 directly, because 
these measurements have been calibrated onto the Lick/IDS sys- 
tem by correcting for Lick offsets. S05 do not provide line index 
measurements. Hence we measure line strengths of all 25 Lick 
absorption-line indices for both sampl es directly on the g lobular 
cluster spectra using the definitions by iTrager et alj (1998). Both 
globular cluster samples have been flux calibrated, so that no fur- 
ther offsets need to be applied for the comparison with the TMJ 
models. We have smoothed the spectra to Lick spectral resolution 
before the index measurement. Note that the spectral resolutions of 
both samples are below the resolution of the MILES library, so that 
we work with the TMJ models at Lick resolution. 

For the P02 sample we adopt the errors quoted in their paper 
using the quadratic sum of the statistical (Poisson) error, the statisti- 
cal error derived from slit to slit variations, and the systematic error 
introduced through uncertainties in the radial velocity. This infor- 
mation is not directly available from S05. We therefore evaluate 
the measurement errors in two steps. First we compute the Poisson 
errors from the error spectra provided through Monte Carlo simu- 
lations. Then we scale these errors with the complete errors from 
P02 from the overlapping globular clusters. 

We add the slit-to-slit error evaluated in P02 in order to ac- 
count for possible stellar population fluctuations that are not in- 
cluded in the statistical error. The observing strategies in both P02 
and S05 have been designed to minimise such an error. Still, this ef- 
fect may not negligible. The slit-to-slit variations overestimate this 
effect, as P02 have typically observed three slits per cluster. We re- 
gard the errors used in this study therefore as conservative estimate, 
and true errors are likely to be smaller. 

Finally, it should be noted that the S05 spectra are corrupt 
around 4546 and 5050 A, so that the indices Fe4531 and Fe5015 
cannot be measured (S05). In case of multiple observations in S05 
we use the spectra with the highest signal-to-noise ratio. 



3 THE TMJ MODEL 

In Paper I we present new stellar population models of Lick 
absorption-line indices with variable element abundance ratios 
(TMJ). The model is an extension of the TMB/K model, which 
is based on the evolut ionary stellar population synthesis code of 
iMarastor] (l998, 2005 ). For basic informatio n on the model we re- 
fer the reader to lThomas et alj (2003a, 2004) and Paper I. Here we 
provide a brief summary of the main features of our new models. 



3.1 New features 

The key novelty compared to our previous models is that the 
TMJ model is flux-calibrated, hence not tied anymore to the 
Lick/IDS system. This is because the new models are based on 
our calibrations of absorption-line indices with stellar parameters 



Chemical abundance ratios of galactic globular clusters 3 



(Joha nsson et alj 2010h derived from the flux -calibrated stellar li- 
brarv MILES dSanchez-Blazquez et al.ll2006h . The MILES library 
consists of 985 stars selected to produce a sample with extensive 
stellar parameter coverage. Most importantly it has been carefully 
flux-calibrated, making standard star-derived offsets unnecessary. 

A further new feature is that we provide model predictions for 
both the original Lick (~ 8 A) and the higher MILES (~ 2.7 A) 
spectral resolutions. Note that the latter appears to be comparable 
to the SDSS resolution, so that our new high-resolution models can 
be applied to SDSS data without any corrections for instrumental 
spectral resolution (see Paper I). 

As a further novelty we calculate statistical errors in the model 
predictions. The errors estimates are obtained from the uncertain- 
ties in the measurements of Lick index strengths and the stellar 
parameters of the library stars, hence do not include systematic er- 
rors. It turns out that the model errors are generally very small and 
well below the observational errors around solar metallicity, but 
rise considerably toward the highest and lowest metallicities. 

The data release now provides models wi th two different stel- 
lar evolutionary tracks by Cassisi et al.1 j 1997b as used in TMB/K 
and additionally Padova dGirardi et alJl200Ch at high metallicities. 
The model based on the Padova tracks is consistent with the model 
using Cassisi for the majority of indices. The cases of indices 
for which the discrepancy exceeds the model error significantly 
are H/3, CNi, CN2, and C24668. Small deviations are found for 
Ca4227, G4300, Ca4455, Fe5015, and Fe5709. In all cases, equally 
for the Balmer and the metal lines, the Padova based models pro- 
duce lower index strengths. This effects kicks in at super-solar 
metallicities, however, outside the range of globular cluster metal- 
licities. In the present study we use the models based on the Cassisi 
tracks. Finally, most importantly for the present paper, we release 
additional model tables with enhancement of each of the elements 
C, N, Na, Mg, Si, Ca, Ti, and Cr separately by 0.3 dex. A differen- 
tial element ratio bias at low metallicities is considered to account 
of the fact that heavier a elements like Ca and Ti tend to be less 
enhanced (see Section|6]l. 

The basic tests with globular cluster data following our strat- 
egy for the TMB/K model is presented in Paper I. The match to the 
globular cluster data is satisfactory for the Balmer line indices H<5a, 
H7A, and H7F. A reasonably good match with globular cluster data 
is also seen for the a/Fe sensitive, metallic indices G4300, Mg 2 , 
and Mg6 and the Fe indices Fe4383, Fe4531, Fe5270, Fe5335, 
Fe5406. The models are well off, instead, for the indices CNi, 
CN 2 , Ca4227, C 2 4668, and Mg 1 . This is caused by the variation 
of further chemical elements beyond the a/Fe ratio to which these 
indices are sensitive, and the full analysis of these element abun- 
dance variations is subject of the present work. In the following 
section we describe how individual element ratios are derived from 
this set of absorption features. 

3.2 Index responses 

Fig. shows the response of the 25 Lick indices to individual el- 
ement abundance changes for a 12 Gyr, solar metallicity stellar 
population. The fractional index change is calculated for an en- 
hancement of the respective element by a factor two normalised to 
the typical observation al measurement error for MILES stars from 
Johan sson et al.1 J2010h . The scale on the y-axis is kept fixed for 
all elements, so that the figure allows us to identify easily those 
elements that are best traced by the current set of models. It can 
be seen that the elements C, N, Na, Mg, Ca, Ti, and Fe are best 
accessible. 



The abundance of nitrogen is obtained from the CN indices 
that are also highly sensitive to C abundance. However, this de- 
generacy can be easily broken through other C sensitive indices 
such as C 2 4668 and Mg 1 . The Mg indices Mg 1; Mg 2 , and Mg6 
are very sensitive to Mg abundance. Note, how ever, that all three 
additionally anti-cor relate with Fe abundance (Trag eTet alj|2000l : 
iThomas et al ] |2003ah . Ca can be measured well from Ca4227, ex- 
cept that this particular index is quite weak and requires good data 
quality. Na abundance can be derived quite easily from NaD in prin- 
ciple. However, in practise this is problematic as the stellar compo- 
nent of this absorption feature is highly contaminated by interstellar 
absorption, which makes this ind ex useless and hence Na inaccessi- 
ble at least for globular clusters dThomas et alj2003ah . Iron is well 
sampled through the Fe indices. 

There are two among the Fe indices, however, that are also 
sensitive to Ti abundance besides Fe. These are Fe4531 and 
Fe5015. They offer the opportunity to estimate also Ti abundance. 
We will only use Fe4531, as Fe5015 is contaminated by a non- 
negligible Mg sensitivity besides Fe, which weakens its usefulness 
for Ti abundance determinations. 

The remaining three elements O, Si and Cr cannot easily be 
m easured through the av ailable indices. As discussed extensively 
in IThomas et ail {2003a), however, oxygen has a special role. O 
is by far the most abundant metal and clearly dominates the mass 
budget of 'total metallicity'. Moreover, the a/Fe ratio is actually 
characterised by a depression in Fe abundance relative to all light 
elements (not only the a elements), hence a/Fe reflects the ratio 
between total metallicity to iron ratio rather than a element abun- 
dance to iron. As total metallicity is driven by oxygen abundance, 
the a/Fe derived can be most adequately interpreted as O/Fe ratio. 
We therefore re-name the parameter a/Fe to O/Fe under the as- 
sumption that this ratio provides an indirect measurement of oxy- 
gen abundance. 

Finally, iron abundance can be calculated through the follow- 
ing for mula dTantalo et al.ll 19981 ; iTrager et alj[2000l : IThomas et al.l 
l2003ah . 

[Fe/H] = [Z/H] - 0.93 [a/Fe] = [Z/H] - 0.93 [O/Fe] 



4 METHOD 

A full description of the method we use to derive individual element 
abundances and its application to SDSS galaxy data is presented 
in a companion paper (Johansson et al, in preparation). Here we 
provide a brief summary of the key aspects that are most relevant 
for the present work. 



4.1 The x 2 technique 

The derivation of the above set of element r atios is done in variou s 
iterative steps by means of the \ 2 code of IThomas et al.1 d2010h . 
We generate a fine grid of model predictions for the parameters 
log age, metallicity, and a/Fe ratio with log steps of 0.1, 0.1, and 
0.05 dex, respectively. Galactic globular cluster data generally are 
very close to the 15 Gyr model (see Paper I), which is the high- 
est age for which we have stellar evolutionary track calculations 
available. Therefore, we extrapolate the models logarithmically to 
a maximum age of 20 Gyr for the initial set of templates, in or- 
der not to impose an upper age limit. Note that the index strengths 
evolve very little as a function of age at these old ages, therefore 



4 Thomas, Johansson, Maraston 



ro 
o o 



ro 
o o 



ro 
o o 



Mg 

ro 
o o 



Na 

ro 
o o 



Ca 



Si 



Ti 

ro 
o o 



Fe 

ro 
o o 



Cr 



1 1 1 1 r i 1 1 1 1 i 1 1 1 1 1 



□ 



1 1 1 1 i 1 1 1 



HdA 

HdF 

CN1 

CN2 

Ca4227 

G4300 

HgA 

HgF 

Fe4383 

Ca4455 

Fe4531 

C24668 

Hbeta 

Fe5015 

Mgl 

Mg2 

Mgb 

Fe5270 

Fe5335 

Fe5406 

Fe5709 

Fe5782 

NaD 

TiOl 

Ti02 



Zl 
□ 



Figure 1. Response of the 25 Lick indices to individual element abundance changes for a 12 Gyr, solar metallicity stellar population. The fractional index 
change is calculated for an enhancement of the respective element by a factor of two. The scale on the y-axis is error normalised. The plot range is kept fixed 
for all elements, so that the figure allows us to identify easily those elements that are best traced by the current set of models. 



we do not expect this extrapolation to affect the derivation of indi- 
vidual element abundances significantly. As a sanity check we have 
verified that the globular clusters with ages above 15 Gyr are not 
biased to particular element abundance ratios. 

The code computes the \ 2 between model prediction and ob- 
served index value for all model templates summing over the n 
indices considered: 



X = 



E 



j-obs 



(1) 



The resulting \ 2 distribution is then transformed into a probability 
distribution. By means of the incomplete Y function adopting the 



degrees of freedom as v — nj n di ce 



we compute the proba- 



bility Q that the chi-square should exceed a particular value \ 2 by 
chance. This computed probability gives a quantitative measure for 
the goodness-of-fit of the model. If Q is a very small probability, 
then the apparent discrepancies are unlikely to be chance fluctua- 
tions. 

The solution with the highest Q (i.e. lowest \ 2 ) ls chosen, 
and the 1-er error is adopted from the FWHM of this probability 
distribution. 



4.2 Choice of indices 

Different from iThomas et af. I f2oToh we discard badly calibrated 
indices from the start. In Paper I we find that the set of indices 
that appears to be best calibrated and most suited for the present 
aims are the Balmer line indices H5a, H7A, and H7F, the metallic 
indices CNi, CN 2 , Ca4227, G4300, C 2 4668, Mg 1; Mg 2 , Mg6, 
and the Fe indices Fe4383, Fe4 531, Fe5270, Fe5335, and Fe5406. 
Then, similar to the approach of lGraves & Schiavonl 1 20081 ) we use 
different sets of indices for different elements. 



4.3 Derivation of individual element abundances 

We define a base set of indices including Mg b, the Balmer index 
H<5 A , and the iron indices Fe4383, Fe5270, Fe5335, Fe5406. First 
we determine the traditional light-averaged stellar population pa- 
rameters age, total metallicity, and a/Fe ratio from this base set of 
indices. Only indices that are sensitive to these three parameters are 
included in the base set. In the subsequent steps we add in turn par- 
ticular sets of indices that are sensitive to the element the abundance 
of which we want to determine. In each step we re-run the \ 2 fitting 
code with a new set of models to derive the abundance of this ele- 
ment. This new set of models is a perturbation to the solution found 
for the base set. It is constructed by keeping the stellar population 
parameters age, metallicity, and a/Fe fixed and by modifying the 



Chemical abundance ratios of galactic globular clusters 5 




log Age (CMD) [Z/H] (CMD) 

Figur e 2. Ages of galac tic g lobular clusters derived from integrated light spectroscopy in comparison with literature data. Glob ular cluster spectra are take n 
from iPuzia et alj 120021) and[§chiavon et al] 120051) . Literature ages from colour-magnitude isochrone fitting are adopted from Marin-Franch et al. (2009). 
Left-hand panel: Grey symbols are the full sample, orange and blue symbols are me tal-rich ([Z/Hl > — 0,8 dex) and metal-poor ([Z/H] < —1.55 dex) 
sub-samples, respectively. The dotted line marks the age of the univer s e as d erived in Ko matsu et al moid) . Ri ght-hand panel: Me tallicity versus horizontal 
branch morpholo gy (horizont al branch ratio HBR adopted from lHarrisi 1 1996)). The literature metallicities on the lZinn & West 1 1984) scale are taken from the 
compilation bv lHarrisj ll996l) . The magenta symbols are those clusters for which we obtain ages larger than 14 Gyr, while the cyan symbols are clusters for 
which our ages agree with the CMD ages within 0.1 dex. The dotted lines are the metallicity limits from the left-hand panel. Literature ages are generally well 
reproduced. We tend to over-estimate ages for the most metal-rich globular clusters. Horizontal branch morphology only plays a minor role. 



element abundance of the element under consideration by ±1 dex 
in steps of 0.05 dex around the base value. 

A new best fit model is obtained from the resulting y 2 distri- 
bution. Then we move on to the next element. At the end of the 
sequence we re-determine the overall \ 2 using all indices together 
and re-derive the base parameters age, metallicity, and a/Fe for the 
new set of element ratios. Then we go back to the second step and 
use these new base parameters to derive individual element abun- 
dances. This outer loop is iterated until the final \ 2 stops improving 
by more than 1 per cent. The method converges relatively fast and 
we generally require five steps at most to fulfil this criterion. 

In more detail, the sequence of elements is as follows. The 
first element in the loop is carbon, for which we use the indices 
CNi, CN 2 , Ca4227, H7A, Efyp, G4300, C 2 4668, Mg l5 and Mg 2 
on top of the base set. Next we drop these C-sensitive indices and 
proceed deriving N abundance, for which we use the N-sensitive 
indices CNi, CN2, and Ca4227. Then we move on to Mg 1 and 
Mg 2 for Mg abundance, Ca4227 for Ca, and finally Fe4531 for the 
element Ti. O abundance is indirectly inferred from the a/Fe ratio 
by assuming that 



[O/Fc] EE [q/Fc] 



(2) 



The typical errors for the parameters are 0.165 dex for log 
Age, 0.21 dex for [Z/H], 0.08 dex for [O/Fe], and about 0.15 dex 
for the other element ratios. It should be emphasised again that 
these are very conservative error estimates. 



5 RESULTS 

From the \ 2 fitting as described in Section [3] we obtain age, to- 
tal metallicity [.Z/H], iron abundance [Fe/H], the [a/Fe] ratio, 
and the individual element abundance ratios [C/Fe], [N/Fe], [O/Fe], 
[Mg/Fe], [Ca/Fe], and [Ti/Fe] for a total of 52 globular clusters. 
We exclude the 47 Tucanae from our analysis, as age dating of 
this cluster from B aimer line indices is know n to be problem- 
atic dSchiavon et a l. 2002; Vazdekis et al. 2001). In this section we 



present the results and compare with literature data obtained from 
colour-magnitude (CMD) fitting (age) and high-resolution spec- 
troscopy of individual stars (metallicity and element abundance ra- 
tios). 



5.1 Ages 

First we discuss the comparison of the derived ages with li tera- 
ture data. CMD ages ar e taken from Marin-Franc h et ail d200 9) and 
|Pe Angeli et al.l yOOS) where not available in Marfn-Franch et ail 
d2009h The overlap of the two samples contains 39 clusters. 

The age comparison is shown in the left-hand panel of Fig. [2] 
where we plot our derived ages as a function of the CMD ages. 
The grey symbols are the full sample, orange and blue are metal- 
rich and metal-poor sub-samples, respectively. The age derivation 
through the Lick indices works reasonably well. Almost half of 
the sam ple our globular cluster ages agree with lMann-Franch et al.l 
d2009t) within 0.1 dex (18 out of 39), and three quarters (29) of the 
clusters agree within the (conservative) measurement errors. Most 
importantly, 35 clusters o ut of 52 (two thirds) are younger than 
the universe as derived by I Komatsu et alj d2010l . dotted line) from 
a combination of cosmic microwave background, supernova, and 
large-scale structure data, and the vast majority (45 out of 52) are 
consistent with age of the universe within 0.1 dex. All but one 
cluster ages are consistent with the age of the universe within our 
(conservative) measurement errors. 

Generally, the ages from Lick indices agree well with the 
CMD ages, albeit with quite a large scatter. Note that lMendel et al 
d2007t) derived systematically larger ages with the TMB/K mode]J 
The turnoff brightness, being the major indicator for the age of 



1 Note that we hav e not explicitly derived globular cluster ages in 
iThomas et alj l2 003al). with respect to CMD. T he major difference is 
that lMendel et alj 120071) adopted CMD ages from lDe Angeli et alj 120051) 
who d erived systematically younger absolute ages than Mari n-Franch et alj 
(2009). It should be kept in mind, however, that the derivation of absolute 



6 Thomas, Johansson, Maraston 



a stellar population, is highly sensitive to the distance of the ob- 
ject. As a consequenc e , only relative ages can be reliably measured 
dOrtolani et alj 1 19951 : be Angeli etaD 120051 : iMarin-Franch et al.1 
l2009h . 

Still, it is interesting to investigate the reason for the exceed- 
ingly large ages of some clusters. Fig. [2] shows that the clusters for 
which we overestimate the ages with respect to the age of the uni- 
verse tend to be metal-rich with [Z/~H] > —0.4 dex (orange sym- 
bols). Most of the clusters whose ages agree well with the CMD 
ages, instead, are metal-poor with [Z/H\ < —1.4 dex (blue sym- 
bols). To investigate thi s further, in the right -hand panel of Fig. [2] 
we plot metallicity (on | Zinn & Westl d 19841) scale adopted from 
IMarin-Franch et al.1 d2009) T versus horizontal branch morphology 
expres sed as horizontal branch ratio HBR adopted from iHarrisI 
(1996). The magenta symbols are those clusters for which we ob- 
tain ages larger than 14 Gyr, while the cyan symbols are clusters 
for which our ages agree with the CMD ages within the measure- 
ment errors. It can be seen clearly that horizontal branch morphol- 
ogy only plays a minor role at a given metallicity. Metallicity is the 
main driver for the age discrepancy. To summarise, globular cluster 
ages derived from absorption-line indices tend to be overestimated 
for the most metal-rich clusters around slightly sub-solar metallic- 
ity. This is a more general manifestation of the H/3 anomaly of 
globular cluster data noted bv lPoole et ail [2010). which may well 
be a 'Balmer anomaly' rather than being restricted to H/3. Note, 
however, that this pattern is more likely to be caused by problems 
in the globular cluster data than the models, as gal axy data appear 
to be well matched instead dKuntschner et alj2010l Paper I). 



5.2 Metallicity 

For the comparison of me tallicit y we adopt literature values from 
iMarm-Franch et al.1 j2009h on the lZinn & Westldl984l) scale . These 
represent total metallicity [Z/H] as Marfn-Franch et ail (2009) 
have corrected the iron measurement using the prescription of 
ISalaris et alj dl993l) . We therefore confront these literature values 
with our measurements of total metallicity [Z/H] in Fig. [3] The 
magenta symbols are those clusters for which we obtain ages larger 
than 14 Gyr, while the cyan symbols are clusters for which our ages 
agree with the CMD ages within 0.1 dex. 

Metallicities agree very well, with a tendency of slightly lower 
metallicity estimates from the present work at low metallicities. It 
can further be seen from Fig.[3]that this match is particularly good 
for those clusters whose Lick index ages agree best with the CMD 
ages (cyan symbols). For the clusters with the oldest Lick index 
ages (magenta symbols), instead, we tend to underestimate metal- 
licity by ~ 0.2 dex This might in fact be an artefact produced by 
the age-metallicity degeneracy, i.e. we tend to underestimate the 
metallicity of those globular clusters for which we overestimate the 
age. 



5.3 Element abundance ratio pattern 

We now turn to discuss the individual abundances of the elements 
C, N, O, Mg, Ca, and Ti relative to the abundance of Fe. Fig. [4] 
presents the abundance ratios [C/Fe], [N/Fe], [Mg/Fe], [Ca/Fe], 
and [Ti/Fe] (coloured symbols) as functions of iron abundance 
[Fe/H] in comparison to [O/Fe] (grey symbols), the bottom panels 



globula r cluster ages through C MD s carries its own problems a s pointed out 
in both lDe Angeli et alj 120051) and lMarin-Franch et alj [2009). 



N 




[Z/H] (CMD) 

Figure 3. Total metallicities [Z/H] of galactic globular clusters derived 
from integrated light spectroscopy i n compariso n with litera ture data. Glob- 
ular cl uster spectra are taken from IPuzia et al 1 fe002l) andlSchiavon et alj 
<2005l) . The literature metallicities on the IZinn & Wesi ll984l) scale and 
corrected for o-enhancement are taken from Man'n- Franch et al.1 j2009h . 
The magenta symbols are those clusters for which we obtain ages larger 
than 14 Gyr, while the cyan symbols are clusters for which our ages agree 
with the CMD ages within 0.1 dex. Literature metallicities are well repro- 
duced. Metallicities are slightly underestimated for those clusters for which 
we overestimate the age. 



indicate the typical measurement error as a function of iron abun- 
dance. 

The element ratio [O/Fe], being equivalent to [a/Fe] (see 
equation [2}, carries the smallest measurement error. This element 
ratio is well determined at all metallicities. The expected pattern 
of super-solar [a/Fe] with a slight decrease toward solar metal- 
licity is reproduced. The individual element abundance ratios, in- 
stead, have significantly larger errors. In all cases the typical er- 
rors increase with decreasing metallicity, and exceed ~ 0.1 dex at 
[Fe/H] < — 1 dex. In fact, the abundance pattern loses structure at 
such low iron abundance. This ought to be expected as the sensitiv- 
ity of the models to element ratio variations decreases dramatically 
with decreasing metallicity (e.g.. lThomas et alj |2003a). Moreover, 
model errors become comparable to model variations for only mod- 
erate abundance ratio changes (see Paper I), which further hampers 
the analysis. Hence abundance ratios cannot be reliably determined. 
At [Fe/H] > — 1 dex, instead, our results reveal interesting abun- 
dance trends. 



5.4 Comparison with stellar spectroscopy 

Before discussing these abundance patterns in detail, we present 
the direct c ompa rison of our results with the measurements of 
|PritzletalJ (2005). who have derived element abundance ratios of 
a large sample o f galactic globular clusters form individual stel- 
lar spectroscopy. IPritzl et al] d2005h have observed between one 
and ten stars per cluster and derived the element ratios [Mg/Fe], 
[Ca/Fe], and [Ti/Fe]. The ratio [a/Fe] is computed from the ge- 
ometrical mean of these three measurements. We have computed 
the strai ght average when m ore than one star has been observed. In 
total, the lPritzl et alj J2005h sample has 18 clusters in common with 
the present work. 

In Fig. [5] we plot the abundance ratios [Mg/Fe], [Ca/Fe], 
and [Ti/Fe] (coloured symbols) as derived in the present work 



Chemical abundance ratios of galactic globular clusters 7 




Figure 4. Abundance ratios [C/Fe], [N/Fe], [Mg/Fe], [Ca/Fe], and [Ti/Fe] (coloured symbols) in comparis on to the the [O/ Fe] r atio (grey symbols) a s 
functions of iron abundance [Fe/H] for galactic globular clusters. The globular clust er spectra are taken from lPuzia et alj |2002|) and Schiavon et al. (2005). 
The black dots are the element ratios of globular cluster stars from lPritzl et alj fcoOib. The bottoms panels show the 1-<t error on the element ratios with the 
dotted horizontal lines indicating an error of 0.15 dex. The typical abundance pattern of Milky Way field and globular cluster stars is well reproduced for 
[O/Fe]. The other element ratios have too large errors at low metallicities below [Fe/H] ~ — 1 dex, hence meaningful conclusions can only be drawn at 
[Fe/H] > -1 dex. 



PL, 



! [a/Fe] 
0.6 - [Mg/Fe] 

[Ca/Fe] 

0.4 - tTi/Fe] 



0.2 



-0.2 



-0.4 



0.2 
[X/Fe] 



0.4 

(Pritzl) 



0.6 



Figure 5. Abundance ratios [Mg/Fe], [Ca/Fe], and [Ti/Fe] (coloured sym- 
bols) of galactic globular clusters measured in this work from integrated 
light spectroscopy in comparison with literature values from individual stel- 
lar spectroscopy by Pritzl et al. (2005). Black symbols show [a/Fc] ratios. 
The error symbols indicate typical errors in both [a/Fc] and [Mg/Fe]. The 
small symbols are globular clusters with [Fe/H] < — 1 dex, for which ele- 
ment ratios from integratd light spectroscopy are unreliable (see Fug.|4). 



from integ rated light spectros copy as functions of the measure- 
ments from IPritzl et alj d2005h . Black symbols show [a/Fe] ra- 
tios. The error symbols indicate typical errors in both [a/Fe] and 
[Mg/Fe]. The small coloured symbols are globular clusters with 
[Fe/H] < — 1 dex, for which individual element ratios from in- 
tegrated light spectroscopy are less reliable (see Fig.[4j- There is 
a satisfactory agreement for [a/Fe] at all metallicities, in agree- 



ment with the results of iMendel et all (2007) obtained with the 
TMB/K models. [Mg/Fe] ratios are still in reasonable agreement 
at [Fe/H] > —1 dex. There is a hint for systematically lower 
[Ca/Fe] and [Ti/Fe] ratios in our work, instead. We present a full 
discussion of this discrepancy in the following sections. 



5.5 Element abundance distributions 

In the following we compare the distributions of element ratios 
from the integrated and stellar spectroscopy. We only consider clus- 
ters with [Fe/H] > — 1 dex in our analysis leaving us with a sample 
of 15 objects (out of 52). The reason is that element ratios cannot 
be reliably determined at lower metallicities, because the relative 
sensitivity of the model predictions to element ratio changes is too 
small (see Fig.|4]l. Note also that we only can consider the P02 sam- 
ple for Ti (6 out of 12) as the Ti sensitive index Fe4531 cannot be 
measured for the S05 clusters. 

Fig. [6] shows the distributions of [C/Fe], [N/Fe], [Mg/Fe], 
[Ca/Fe], and [Ti/Fe] ratios (coloured histograms) in comparison 
with the distribution of the [O/Fe] ratio (grey histograms). The 
median values of these distribution are given in Table [TJ The dis- 
tribution of [O/Fe] is reasonably tight with a median value of 
0.24 dex as expected for Milky Way globular clusters. The other of 
the light a elements considered, [Mg/Fe], follows this distribution 
very closely with a very similar median of 0.25 dex. The other light 
elements, instead, deviate from this pattern. [C/Fe] and [Ca/Fe] ra- 
tios show similarly peaked distributions, but with different median 
values. The element ratio [C/Fe] has a median slightly lower by 
~ 0.04 dex, while [Ca/Fe] peaks at significantly lower values with 
a median of 0.15 dex. A Kolmogorov-Smirnov test confirms that 
the distributions in [O /Fc] and [Ca/Fe] come from different under- 



8 Thomas, Johansson, Maraston 




[O/Fe] 
[C/Fe] 



0.5 
[X/Fe] 



1 1 1 1 1 1 1 1 


1 1 1 1 1 1 1 

[O/Fe] - 




[Mg/Fe] - 






Pritzl 









0.5 

[X/Fe] 




1 1 1 1 1 1 1 1 


1 1 1 1 1 1 1 

[O/Fe] - 

- [Ca/Fe] 
Pritzl 


III 





0.5 

[X/Fe] 




Figure 6. Distributions of [C/Fe], [N/Fe], [Mg/Fe], [Ca/Fe], and [Ti/Fe] ratios (coloured histograms) in compari son to the distribution of the [O/Fc] ratio 
(grey histogram) for galactic globular clusters. The globular cluster spectra are taken from Puzia et al. (2002) and Schi avon et alj J2005t) - Only clusters with 
[Fe/H] > — 1 dex are considered (15 objects out of 52) as element ratios cannot be reliably determined at lower metallicities (see text). The dotted black lines 
(for [Mg/Fe], [Ca/Fe], [Ti/Fe] only) are the distributions of the element ratios of globular cluster stars from IPritzl et alj 120051) . We find a general trend such 
that the heavier of the light elements (Ca and Ti) are less enhanced than O and Mg. N is strongly enhanced in a sub-population of clusters accompanied by a 
slight depression of [C/Fe] with respect to O and Mg. 



Table 1. Median values of element ratio distributions (in dex). [Ti/Fe] ratio is less well determined, but the results suggest that 

this element continues this trend with even lower [Ti/Fe] ratios. 



[O/Fe] 


[C/Fe] 


[N/Fe] 


[Mg/Fe] 


[Ca/Fe] 


[Ti/Fe] 


0.24 


0.20 


0.71 


0.25 


0.15 


0.09 



lying distributions at the > 6a level. The distribution of [N/Fe] has 
a pronounced peak at a significantly larger value of 0.71 dex. The 
distribution of [Ti/Fe] is somewhat scattered. Still, the data show a 
clear trend toward lower [Ti/Fe] ratios with a median of 0.09 dex 
in line with the neighbouring a element Ca. 

The significant enhancement of nitrogen together with the 
slight depression of carbon relative to the other light elements 
is a well known abundance pattern in globular clusters observed 
in high-resolution spectroscopy studie s of individual stars (e.g., 
iNorris etal.ll 1984 ICarretta et all2005t) . This chemical anomaly is 
commonly a ttributed to self-enric hment during the formation of the 
star cluster ( Ventura et al. 2009). Such N enhancement has been 
quantified in lThomas et al. (2003a) for the first time for integrated 
light observations of globular clusters, while the accompanying de- 
pression of C found in the present work is new. The [C/Fe] and 
[N/Fe] ratios deriv ed here appear to be well consistent with the 
measurements from lCarretta et alj fe005l) . 

The next heavier of the light elements, Mg, follows the distri- 
bution of [O/Fe] closely. This ought to be expected as these el- 
ements are close in atomic number and create d in very similar 
processes during supernova n ucleosynthesis dWooslev & Weaveil 
Il995t iThielemann et alj|l99d) . However, the heavier a elements 
Ca and Ti deviate from this pattern. The typical [Ca/Fe] ratio is 
significantly lower than the typical [O/Fe] and [Mg/Fe] ratios. The 



5.6 New model fits 

The adjustment of individual element abundances helps to im- 
prove the fits to a number of indices. In Figs. [7] and [8] we 
revisit the calibration figure from Paper I for the model af- 
ter the full chemical analysis. We only show those indices that 
have been used in the analysis, plotting index strengths as func- 
tions of [MgFe]'. Three models at Lick spectral resolution with 
an age of 13 Gyr are shown for the metallicities [Z/H] = 
-2.25, -1.35, -0.33, 0.0, 0.35, 0.67 dex. Metallicity increases 
from left to right. The solid lines are the final model for the av- 
erage of the individual element abundance ratios derived through 
the x 2 fit- The dotted and dashed lines are the base models with 
a/Fe = 0.0 dex and a/Fe = 0.3 dex for comparison. The 
grey shaded area along the model indicates the 1-ct error of the 
model prediction. Galactic globular clusters from P02 and S05 
are filled and open squares, respectively. The typical errors in the 
globular cluster index measurements are given the error symbol 
at the bottom of each panel. The small black dots are early-type 
galaxies from the MOSES catalogue (Morphologically Selected 
Early -type galaxies in SDSS Schawin ski et alj2007l ; lThomas et all 
l2010h drawn from the SDSS (Sloan Digital Sky Survey) data base 
I York et alj|2000h including only high signal-to-noise spectra with 
S/N > 40. 



Chemical abundance ratios of galactic globular clusters 9 



z 

u 



0.2 h 
0.1 


-0.1 



r- 
w 

CO 

O 



CO 

co 

CO 

w 
O 



10 - 




0.5 
0.4 
0.3 
0.2 
0.1 





1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1, < 1 
N enhanced s 


1 1 1 1 | 1 1 1 1 | 1 1 1 1 | 1 1 1 1 | 1 1 1 1 | 1 1 


1 1 1 I | 1 1 1 1 | 1 I 1 1 | 1 I 1 I | 1 1 1 1 1 1 I 
Ca depressed f 


l l l l | l l l l | l l l l | l l l l | l l l l | l l" 
C depressed 

+ ; 


1 1 1 1 | 1 1 1 1 | 1 1 1 1 | 1 1 1 1 | 1 1 1 1 | 1 1 
C depressed 

--^^^^^^ + 


1 1 1 1 | 1 1 1 1 | 1 1 1 1 | 1 1 1 1 | 1 1 1 1 | 1 1 
Mg enhanced 


_l l l l | l l l l | l l l l | l l l l | l l l l | l l 
Mg enhanced 

C depressed ^^HL 
: **** + 

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 L 


1 1 1 1 | 1 1 1 1 | 1 1 1 1 | 1 1 1 1 | 1 1 1 1 | 1 1 
Mg enhanced 

+ " 



-0.1 



o 

CO 

o 
o 



- 0.2 



g, 

0.1 2° 



4 2 

00 

Cr 



- 2 



1 2 3 4 5 
[MgFe]' 



1 2 3 4 5 
[MgFe]' 



Figure 7. Calibration of the line indices that are sensitive to light elements. Three models at Lick spectral resolution with an age of 13 and the metallicities 
[Z/H] = —2.25, —1.35, —0.33, 0.0, 0.35, 0.67 dex are shown. The solid lines are the final model for the average of the individual element abundance 
ratios derived through the x 2 fit- The dotted and dashed lines are the base models with [a/ Fe] = 0.0 dex and [q/ Fe] = 0.3 dex. The gre y shaded area along 
the model indicates the 1-tr error of the model prediction. Galactic globular clusters from lPuzia et al ] j2002h and lSchiavon et al l <2005l) are filled and open 
squares, respectively. The typical errors in the globular cluster index measurements are given the error symbol at the bottom of each panel. The small black 
dots are early-type galaxies from the MOSES catalogue (Morphologically Selected Early-type galaxies in SDSS[Schawinski et al. 2007; Thomas et alj20ld) 
drawn from the SDSS (Sloan Digital Sky Survey) data base jYork et alj2000h including only high signal-to-noise spectra with S/N > 40. 



5.6.1 Light element indices 

Fig. [7] shows the indices that are sensitive to light element abun- 
dances, namely CNi, CN 2 , Ca4227, G4300, C 2 4668, M gl , Mg 2 , 
and Mg b. It can be seen from the top panels that the strengths of 
the CN indices are clearly underestimated in both solar-scaled and 
a/Fe enhan ced base models (dott ed and dashed lines). As already 
discussed in lThomas et alj d2003ah a significant enhancement in N 
is required to explain the high index strengths. At the same time, 



the indices G4300, and C24668, and Mgj^ are slightly too strong in 
the base models, which leads to a slight reduction of C abundance 
in the final best-fitting model. 

Another striking element abundance pattern that can be in- 
ferred from Fig.|7]directly is the abundance of Ca. The strength of 
the index Ca4227 is significantly over-predicted by the solar-scaled 
and a/Fe enhanced base models. The model matches the globular 
cluster data very well, instead, when a depression of Ca abundance 
is included. Finally, the indices Mg 2 and Mg b (bottom panels) are 



10 Thomas, Johansson, Maraston 



X 



CO 
CO 



co 

CO 
lO 
0) 

fe. 



5 - 



- 



4 - 



2 - 



i i i i — I ii — i i I i — i r 



i i i i I — r 



Fe depressed 



Kin 




I I I I | I I I I | I I I I | I I I I | I I I I | I I 

Fe depressed 




l l l l | l l l l | l l l l | l l l l | l l l I | l I 

Fe depressed 
Ti depressed 



I I I I | I I I I | I I I I | I I I I | I I I I | I I 



Fe depressed 




I I I I I I I I I — I — TH — I I I I 



"T 




t 

Fe depressed 




- -5 



I I I I | I I I I | I I I I | I I I I | I I I I | I I 
Fe depressed j 




- 2 



I I I I | I I I I | I I I I | I I I I | I I I I | M 



Fe depressed 




l l l l | l l l l | l l l l | l l l l | l l l l | l I 



Fe depressed 









+ 



■10 



- 4 



- 3 



- 



1 2 3 4 5 
[MgFe]' 



[MgFe]' 



CD 

w 

CD 
CO 



a 
en 
ro 

o 



CD 

O 

03 



Figure 8. Calibration of the Fe and Balmer line indices. Three models at Lick spectral resolution with an age of 13 and the metallicities [Z/B] = 
—2.25, —1.35, —0.33, 0.0, 0.35, 0.67 dex are shown. The solid lines are the final model for the average of the individual element abundance ratios 
derived through the \ 2 fit. The dotted and dashed lines are the base models with [a/Fe] = 0.0 dex and [q /Fe l = 0.3 dex. The gre y shaded area along 
the model indicates the l-cr error of the model prediction. Galactic globular clusters from Puzia et al. 12002) and Schi avon et al . (2005) are filled and open 
squares, respectively. Note that the indices Fe4531 and Fe5015 cannot be measured for the lSctu^vonetalJ 120051) sample. The typical errors in the globular 
cluster index measurements are given the error symbol a t the bottom of each panel. The small black dots are early-type galaxies from the MOSES catalogue 
(Morpholog ically Sel ected Early-type galaxies in SDSS Schawins ki et al ] |2007l ; lThomas et alj|2010l) drawn from the SDSS (Sloan Digital Sky Survey) data 
base jYork et al.l2OO0l) including only high signal-to-noise spectra with S/N > 40. 



well reproduced by the a/Fe enhanced model (dashed line), and 
only a negligible adjustment of Mg abundance is required to opti- 
mise the fit. 



5.6.2 Iron and Balmer line indices 

Fig. [8] presents the Fe and Balmer line indices used in the fitting 
procedure. The solar-scaled model (dotted lines) generally over- 



predicts the index strengths of the Fe indices, which is remedied 
through a depression of Fe abundance in the a/Fe enhanced model 
(dashed lines). The index strength of Fe4531 is slightly re-adjusted 
through a depression of Ti abundance. The signal is very weak, 
though, and the determination of Ti abundance in this work must 
in fact be considered tentative, in particular since only a handful of 
clusters from P02 are available for the Ti abundance measurement. 
In general, the full chemical model only leads to minor corrections 



Chemical abundance ratios of galactic globular clusters 1 1 



of the Fe indices. The same is true for the Balmer line indices. 
Here the solar-scaled model under-predicts line strengths, which 
is remedied by the enhancement of the a/Fe ratio. Again, other 
elements only have negligible impact on these indices. 



els of bulges and spheroids predict lower element abundances fo r 
the heavier a elements, Ca in particular dMatteucci et al.|[l999h . 
This implies that also Ti would have to be underabundant in mas- 
sive galaxies. We investigate this issue in a companion paper (Jo- 
hansson et al, in preparation). 



6 DISCUSSION 

We have derived, for the first time, detailed chemical element abun- 
dance patterns of galactic globular clusters from integrated light 
spectroscopy. The light elements O and Mg show the well-known 
enhancement with respect to Fe, hence [O/Fe] ~ [Mg/Fe] ~ 
0.3 dex. For C, N, and the heavier a elements Ca and Ti, how- 
ever, we detected interesting abundance anomalies. N is further en- 
hanced to very high [N/Fe] ratios, while C is slightly depressed. 
Ca exhibits significantly lower [Ca/Fe] ratios than O or Mg, a pat- 
tern that appears to be present also in [Ti/Fe]. These anomalies 
have interesting consequences for supernova nucleosynthesis and 
the chemical enrichment in the Milky Way. 

First we confront these results with the ele ment ratios of indi - 
vidual stars in globular clusters as measured bv |Pritzletai] f2005). 
These are shown by the dotted lines in Fig.[6]for [Mg/Fe], [Ca/Fe], 
and [Ti/Fe]. The distributions of the [Mg/Fe] ratios agree very 
well. [Ca/Fe] and [Ti/Fe] ratios, instead, are somewhat higher in 
IPritzl et al] d2005h . The latter do suggest slightly lower enhance- 
ment of these elements, but not as pronounc ed as found here. Bu t 
our finding gets support from the study by iFeltzing et al.l ( 2009) 
who analyse six horizontal branch stars in the metal-rich galactic 
globular cluster NGC 6352. As expected the cluster is enhanced in 
the a-elements. But like in the present work lFeltzing et al.l 1 2009. 
see their Fig. 7) find a sequence of decreasing element ratios rela- 
tive to iron for increasing atomic numbers from Mg through Ca to 
Ti. 

It should be expected that field stars show the same behaviour 
since globular cluster element abundances generally follow the 
ones of the field stars in the galactic halo and discs jPritzl et al] 
2005). In fact the trend reported here starts to crystallise out now 
from recent high quality stellar spectr oscopy of galactic field stars 
in bulge and disc. lBensbv et analyse bulge and thick/thin 

disc stars and find [Ca/Fe] and [Ti/Fe] ratios to be lower than [O/Fe] 
and [Mg/Fe] ratios for all three populations. 

The implication is that some fraction of the abundance in the 
heavier a elements must come from Type la supernova explosions, 
while the lighter elements O and Mg remain to be enriched exclu- 
sively by Type II. The yields of the W7 model for Type la supernova 
explosions do indeed predict the production of traces of the heaviest 
a elements. As a consequence, galactic che mical evolution model s 
predict lower [Ca/Fe] ratios for halo stars dChiappini et alj| 19971) . 
which had not been confirmed from observational data so far. The 
results discussed here provide a new observational support for this 
pattern. 

This is critical for the chemical enrichment histories of galax- 
ies. It leads to the most natural explanation for the shallow 
slop e of the [Ca/Fe] - galaxy mass relation of early-type galax- 
ies dSaglia et alJ|2002| : [Thomas et ai] l2003bl : ICenarro et ail [20031 : 
Michi elsen et al]|2003l) . In this scenario, Ca is underabundant rela- 
tive to the lighter a elements in massive gal axies for the same rea - 
son as Fe is underabundant (see discussion in lThomas e t al. 2003b). 
The short formation time-scales inhibit Type la supernovae to play 
a role in the chemical enrichment history of the stellar populations 
in these galaxies, such that elements that are produced in Type la 
supernova are depleted in the stars. In fact chemical evolution mod- 



7 CONCLUSIONS 

Modelling integrated light spectroscopy of unresolved stellar pop- 
ulations allows us to study the detailed element abundances in dis- 
tant galaxies and globular clusters. In Paper I we present new, flux- 
calibrated stellar population models of Lick absorption-line indices 
with variable element abundance ratios (TMJ models). The new 
model includes a large variety of individual element variations, 
which allows the derivation of the abundances for the elements C, 
N, O, Mg, Ca, Ti, and Fe besides total metallicity and age. In the 
present paper we use this model to obtain estimates of these pa- 
rameters and element abundance ratios from integrated light spec- 
troscopy of galactic globular clusters. 

The globu l ar clus ter data is taken from Puzia et al] d2002l) and 
ISchiavon et ail d2005h . We measure line strengths of all 25 Lick 
absorption-line indices for both samples directly on the globular 
cluster spectra. Both globular cluster samples are flux calibrated, 
so that no further offsets need to be applied for the comparison 
with the TMJ models. 

We derive the element abundance ratios [C/Fe], [N/Fe], 
[O/Fe], [Mg/Fe], [Ca/Fe], [Ti/Fe] through an iterative x 2 fit- 
ting technique. First we determine the traditional light-averaged 
stellar population parameters age, total metallicity, and a/Fe ratio 
from the indices Mg b, the Balmer index H<5a, and the iron indices 
Fe4383, Fe5270, Fe5335, and Fe5406. In the subsequent steps we 
add in turn particular sets of indices that are sensitive to the element 
the abundance of which we want to determine. The indices used are 
CNi, CN 2 , Ca4227, H 7A , H7F, G4300, C 2 4668, Mg x , and Mg 2 
for carbon, CNi, CN 2 , and Ca4227 for nitrogen, Mg 1 and Mg 2 
for magnesium, Ca4227 for calcium, and Fe4531 for titanium. The 
Ti sensitivity of Fe4531 is relatively weak, hence the abundance 
derivations for this element are only tentative. The [O/Fe] ratio 
is indirectly inferred by assuming that [O/Fe] = [a/Fe]. We show 
that the model fits to these indices in globular clusters improve con- 
siderably through this full chemical analysis. 

The ages we derive agree well with the literature. In particular 
the ages derived here are all consistent with the age of the universe 
within the measurement errors. There is a considerable scatter in 
the ages, though, and we overestimate the ages preferentially for 
the metal-rich globular clusters, which appears to extend the previ- 
ously reported H/3 anomaly of globular clusters to the other Balmer 
indices. Our derived total m etallicities agree gen erally very well 
with literature values on the Izinn & West! (1984) scale once cor- 
rected for a-enhancement, in particular for those cluster where the 
ages agree with the CMD ages. We tend to slightly underestimate 
the metallicity for those clusters where we overestimate the age, in 
line with the age-metallicity degeneracy. 

It turns out that the derivation of individual element abundance 
ratios is highly unreliable at [Fe/H] < — 1 dex, while the [a/Fe] 
ratio is robust at all metallicities. The discussion of individual ele- 
ment ratios focuses therefore on globular clusters with iron abun- 
dances [Fe/H] > — 1 dex. We find general enhancement of light 
and a elements as expected with significant variations for some el- 
ements. The elements O and Mg follow the same general enhance- 
ment with almost identical distributions of [O/Fe] and [Mg/Fe]. 



12 Thomas, Johansson, Maraston 



We find slightly lower [C/Fe] and very high [N/Fe] ratios, instead. 
Hence N is significantly enhanced and C slightly depressed in glob- 
ular clusters with respect to the other light elements. This chemical 
anomaly commonly attributed to self-enrichment is well known in 
globular clusters from individual stellar spectroscopy, and it is the 
first time that this pattern is derived also from the integrated light. 

The a elements follow a pattern such that the elements with 
higher atomic number, namely Ca and Ti, are less enhanced. More 
specifically, [Ca/Fe] ratios are lower than [O/Fe] and [Mg/Fe] by 
about 0.2 dex. Ti continues this trend. We compare this result with 
recent determinations of element abundances in globular cluster 
and field stars of the Milky Way. We come to the conclusion that 
this pattern is now universally found. It suggests that Type la su- 
pernovae contribute significantly to the enrichment of the heavier 
a elements as predicted in supernova explosion calculations and 
galactic chemical evolution models. This explains the presence of 
a Ca under- abundance (close to solar [Ca/Fe] ratios) in massive 
early-type galaxies and predicts similarly low [Ti/Fe] ratios in pop- 
ulations with short formation time-scales. 

Tables with the absorption line indices, stellar population pa- 
rameters and chemical element ratios of the globular clusters anal- 
ysed here are available at www.icg.port.ac.uk/~thomasd. 



ACKNOWLEDGEMENTS 

We would like to thank Harald Kuntschner, Ricardo Schiavon, and 
Gustav Stromback for the many stimulating discussions and for 
providing us with their galaxy and model data. We are deeply in 
dept with Alvio Renzini who pointed us to a mistake with the 
adopted CMD ages in the previous version. We thank the anony- 
mous referee for very valuable comments that helped to improve 
the manuscript. CM acknowldges support by the Marie Curie Ex- 
cellence Team Grant MEXT-CT-2006-042754 of the Training and 
Mobility of Researchers programme financed by the European 
Community. 

Funding for the SDSS and SDSS-II has been provided by 
the Alfred P, Sloan Foundation, the Participating Institutions, the 
National Science Foundation, the U.S. Department of Energy, 
the National Aeronautics and Space Administration, the Japanese 
Monbukagakusho, the Max Planck Society, and the Higher Ed- 
ucation Funding Council for England. The SDSS Web Site is 
http://www.sdss.org/ 

The SDSS is managed by the Astrophysical Research Con- 
sortium for the Participating Institutions. The Participating Institu- 
tions are the American Museum of Natural History, Astrophysical 
Institute Potsdam, University of Basel, University of Cambridge, 
Case Western Reserve University, University of Chicago, Drexel 
University, Fermilab, the Institute for Advanced Study, the Japan 
Participation Group, Johns Hopkins University, the Joint Institute 
for Nuclear Astrophysics, the Kavli Institute for Particle Astro- 
physics and Cosmology, the Korean Scientist Group, the Chinese 
Academy of Sciences (LAMOST), Los Alamos National Labora- 
tory, the Max-Planck-Institute for Astronomy (MPIA), the Max- 
Planck-Institute for Astrophysics (MPA), New Mexico State Uni- 
versity, Ohio State University, University of Pittsburgh, University 
of Portsmouth, Princeton University, the United States Naval Ob- 
servatory, and the University of Washington. 



REFERENCES 

Bensby T, Feltzing S., Johnson J. A., Gould A., Aden D., Asplund 
M., Melendez J., Gal- Yam A., Lucatello S., Sana H., Sumi T, 
Miyake N., Suzuki D., Han C, Bond L, Udalski A., 2010, A&A, 
512 

Burstein D., Faber S. M., Gaskell C. M., Krumm N., 1984, ApJ, 
287, 586 

Carretta E., Gratton R. G, Lucatello S., Bragaglia A., Bonifacio 

P., 2005, A&A, 433, 597 
Cassisi S., Castellani M., Castellani V., 1997, A&A, 317, 10 
Cenarro A. J., Gorgas J., Vazdekis A., Cardiel N., Peletier R. F, 

2003, mnras, 339, L12 
Chiappini C, Matteucci F, Gratton R. G, 1997, ApJ, 477, 765 
Clemens M. S., Bressan A., Nikolic B., Alexander P., Annibali F, 

Rampazzo R„ 2006, MNRAS, 370, 702 
Cohen J.G, 1983, ApJ, 270, 654 

De Angeli F, Piotto G, Cassisi S., Busso G, Recio-Blanco A., 
Salaris M., Aparicio A., Rosenberg A., 2005, AJ, 130, 1 16 

Faber S. M„ Friel E. D., Burstein D., Gaskell D. M„ 1985, ApJS, 
57, 711 

Feltzing S., Primas F, Johnson R. A., 2009, A&A, 493, 913 
Girardi L., Bressan A., Bertelli G, Chiosi C, 2000, A&AS, 141, 
371 

Gorgas J., Faber S. M., Burstein D., Gonzalez J. J., Courteau S., 

ProsserC, 1993, ApJS, 86, 153 
Graves G. J., Schiavon R. P., 2008, ApJS, 177, 446 
Harris W. E., 1996, AJ, 112, 1487 

Johansson J., Thomas D., Maraston C, 2010, MNRAS, in press 
Kelson D. D., Illingworth G. D., Franx M., van Dokkum P. G, 

2006, ApJ, 653, 159 
Komatsu E., et al., 2010, ApJS 

Korn A., Maraston C, Thomas D., 2005, A&A, 438, 685 
Kuntschner H., Emsellem E., Bacon R., Bureau M., Cappellari 
M., Davies R. L., de Zeeuw P. T, Falcon-Barroso J., Krajnovic 
D., McDermid R. M., Peletier R. F, Sarzi M., 2010, MNRAS, in 
press 

Lee H., Worthey G, Dotter A., Chaboyer B., Jevremovic D., 
Baron E., Briley M. M., Ferguson J. W., Coelho P., Trager S. C, 
2009, ApJ, 694, 902 

McWilliam A., 1997, ARA&A, 35, 503 

Maraston C, 1998, MNRAS, 300, 872 

Maraston C, 2005, MNRAS, 362, 799 

Maraston C, Greggio L., Renzini A., Ortolani S., Saglia R. P., 

Puzia T, Kissler-Patig M., 2003, A&A, 400, 823 
Marm-Franch A., et al., 2009, ApJ, 694, 1498 
Matteucci F, Romano D., Molaro P., 1999, A&A, 341, 458 
Mendel J. T, Proctor R. N., Forbes D. A., 2007, MNRAS, 379, 
1618 

Michielsen D., De Rijcke S., Dejonghe H., Zeilinger W. W., Hau 

G. K. T, 2003, ApJ, 597, L21 
Norris J., Freeman K. C, Da Costa G. S., 1984, ApJ, 277, 615 
Ortolani S., Renzini A., Gilmozzi R., Marconi G, Barbuy B., Bica 

E„ Rich R. M., 1995, Nature, 377, 701 
Poole V., Worthey G, Lee H, Serven J., 2010, AJ, 139, 809 
Pritzl B. J., Venn K. A., Irwin M., 2005, AJ, 130, 2140 
Puzia T, Saglia R. P., Kissler-Patig M., Maraston C, Greggio L., 

Renzini A., Ortolani S., 2002, A&A, 395, 45 
Renzini A., 1998, AJ, 115, 2459 

Saglia R. P., Maraston C, Thomas D., Bender R., Colless M., 

2002, ApJ, 579, L13 
Salaris M., Chieffi A., Straniero O. 1993, ApJ, 414, 580 



Chemical abundance ratios of galactic globular clusters 



Sanchez-Blazquez P., Peletier R. R, Jimenez-Vicente J., Cardiel 

N., Cenarro A. J., Falcon-Barroso J., Gorgas J., Selam S., 

Vazdekis A., 2006, MNRAS, 371, 703 
Schawinski K., Thomas D., Sarzi M., Maraston C, Kaviraj S., Joo 

S.-J., Yi S. K., Silk J., 2007, MNRAS, 382, 1415 
Schiavon R. P., Faber S. M., Rose J. A., Castilho B. V., 2002, ApJ, 

580, 873 

Schiavon R. P., Rose J. A., Courteau S., MacArthur L. A., 2005, 

ApJS, 160, 163 
Serven J., Worthey G., Briley M. M., 2005, ApJ, 627, 754 
Smith R. J., Lucey J. R., Hudson M. J., Bridges T. J., 2009, MN- 
RAS, 398, 119 
Tantalo R., Chiosi C, Bressan A., 1998, A&A, 333, 419 
Thielemann F.-K., Nomoto K., Hashimoto M., 1996, ApJ, 460, 
408 

Thomas D., Maraston C, Bender R., 2003a, MNRAS, 339, 897 
Thomas D., Maraston C, Bender R., 2003b, MNRAS, 343, 279 
Thomas D., Maraston C, Johansson J., 2010, MNRAS, submitted 
(Paper I) 

Thomas D., Maraston C, Korn A., 2004, MNRAS, 351, L19 
Thomas D., Maraston C., Schawinski K., Sarzi M., Silk J., 2010, 

MNRAS, 404, 1775 
Tolstoy E., Hill V., Tosi M., 2009, ARA&A, 47, 371 
Trager S. C, Faber S. M., Worthey G, Gonzalez J. J., 2000, AJ, 

119, 164 

Trager S. C, Worthey G, Faber S. M., Burstein D., Gonzalez J. J., 

1998, ApJS, 116, 1 
Tripicco M. J., Bell R. A., 1995, AJ, 1 10, 3035 
Vazdekis A., Salaris M., Arimoto N., Rose J. A., 2001, ApJ, 549, 

274 

Ventura P., Caloi V., D'Antona F, Ferguson J., Milone A., Piotto 

G. P., 2009, MNRAS, 399, 934 
Woosley S. E., Weaver T. A., 1995, ApJS, 101, 181 
Worthey G, Faber S. M., Gonzalez J. J., Burstein D., 1994, ApJS, 

94, 687 

Worthey G., Ottaviani D. L., 1997, ApJS, 1 1 1, 377 
York D. G, et al., 2000, AJ, 120, 1579 
Zinn R., West M. J., 1984, ApJS, 55, 45 

This paper has been typeset from a TgX/ I5TjX file prepared by the 
author.